Networked sensor data error estimation
Networked sensor data error estimation
复制标题
DOI：
10.1016/j.trb.2019.01.013
复制
发表时间：
2019-04
期刊：
影响因子：
--
通讯作者：
Yudi Yang;Han Yang;Yueyue Fan
中科院分区：
文献类型：
--
作者：
Yudi Yang;Han Yang;Yueyue Fan
如今，任何智能交通管理或控制策略的有效性都将在很大程度上取决于传感器收集的可靠交通数据。关于传感器数据质量的两个问题受到关注：第一，识别故障传感器的问题;第二，交通流的重建。大多数现有的研究关注识别完全故障的传感器，其数据应该被丢弃。在本文中，我们专注于部分故障传感器，可以提供有价值的信息的错误检测和数据恢复的问题。通过集成传感器测量误差模型和交通网络模型，我们提出了一种基于广义矩量法（GMM）的估计方法，以确定交通传感器在道路网络中的系统和随机误差的参数。所提出的方法允许灵活的数据聚合，改善识别和准确性。利用系统误差和随机误差的估计值对传感器健康状况进行假设检验，并利用观测计数估计真实流量。不同规模的三个网络的例子的结果表明，所提出的方法在各种各样的情况下的适用性。
Nowadays, the effectiveness of any smart transportation management or control strategy would heavily depend on reliable traffic data collected by sensors. Two problems regarding sensor data quality have received attention: first, the problem of identifying malfunctioning sensors; second, reconstruction of traffic flow. Most existing studies concerned about identifying completely malfunctioning sensors whose data should be discarded. In this paper, we focus on the problem of error detection and data recovery of partially malfunctioning sensors that could provide valuable information. By integrating a sensor measurement error model and a transportation network model, we propose a Generalized Method of Moments (GMM) based estimation approach to determine the parameters of systematic and random errors of traffic sensors in a road network. The proposed method allows flexible data aggregation that ameliorates identification and accuracy. The estimates regarding both systematic and random errors are utilized to conduct hypothesis test on sensor health and to estimate true traffic flows with observed counts. The results of three network examples with different scales demonstrate the applicability of the proposed method in a large variety of scenarios.